Elon Musk praises Anthropic's Mythos/Fable, vows not to cut off

Elon Musk wrote, "I was clearly wrong about Anthropic," and vowed he would not shut the AI lab off, after Anthropic signed a May deal to buy 300 megawatts of compute, the entire output of xAI’s Colossus 1 data center, paying $1.25 billion per month through May 2029.

What did Musk say about Anthropic?

Musk called Anthropic the current leader and said he would not cut the company off even as a competitor, arguing that doing so is "not my style." He added that "No company has released a model as good as Mythos/Fable and they will undoubtedly have Mythos 2 ready soon," framing the praise around the Mythos/Fable models and promising no deliberate infrastructure squeeze.

Beyond the praise, Musk pointed to past business decisions as examples of his restraint: Tesla’s 2014 patent stance and opening of its Supercharger network. He also noted SpaceX launches competing satellite systems with no increase in price or unfair terms, and acknowledged he has taken aggressive legal steps in other contexts, for example suing OpenAI.

What is the Anthropic–SpaceX/xAI hosting deal?

Anthropic signed a May agreement to purchase 300 megawatts of compute, the full output of xAI’s Colossus 1 data center near Memphis, Tennessee, and agreed to pay $1.25 billion per month through May 2029, a contract the article describes as worth about $40 billion in revenue for SpaceX’s xAI unit. The story also notes Google rented SpaceX infrastructure through June 2029 at $920 million per month.

As of July 2026, Anthropic is one of SpaceX’s largest customers. The piece also states xAI merged with SpaceX in February, which folded xAI’s Colossus 1 supply into SpaceX’s infrastructure offerings. The scale of the contract gives SpaceX recurring high-value revenue and places Anthropic’s models on infrastructure controlled by Musk’s companies.

Are there risks from hosting and model "distilling"?

Hosting Anthropic’s compute creates contractual and operational frictions and exposures: contractual consequences would follow if SpaceX suddenly cut service, and SpaceX gains visibility into Anthropic’s operations as it supports the lab’s large-scale AI workloads. Musk conceded in his trial against OpenAI that "Generally AI companies distill other AI companies," acknowledging a common competitive tactic of probing rivals’ models. Anthropic in February accused three Chinese model makers of distilling Claude, evidence the company is sensitive to data-leakage and reverse-engineering risks.

The article adds that SpaceX engineers supporting Anthropic could learn how to build for Anthropic’s workloads, an upside for SpaceX even as the hosting relationship raises potential sightlines into Anthropic’s development practices.

Why it matters

The hosting deal ties together two of the biggest commercial forces in AI and infrastructure: Anthropic’s top-market models and SpaceX/xAI’s concentrated compute capacity. The contract’s scale — 300 megawatts and $1.25 billion per month through May 2029 — creates a high-value dependency that shifts incentives for both sides. That mixture of dependence, visibility and contractual protection will shape competitive dynamics among cloud and AI providers.

What to watch

Watch whether Anthropic releases "Mythos 2" as Musk predicted and whether the three-year hosting contract remains stable as it ages toward May 2029. Also monitor any public disputes over data access or distillation practices, and whether other large cloud customers change their posture toward SpaceX/xAI given these commercial arrangements.